One runner, two flavors

The MMGIS test suite has consolidated on Playwright — every spec file, fast or slow, is now executed by @playwright/test. The top-level README.md still references Jest in passing (a leftover from the v2.0.0 era when Jest was added), but tests/ itself has no Jest config or specs left. That keeps the dependency surface small and lets unit and e2e tests share the same imports, assertions, and HTML reporter.

The split is purely structural:

tests/
├── unit/        # Pure JS, no browser. Imports modules directly.
├── e2e/         # Real Chromium, real backend, navigates to '/'.
├── fixtures/    # Static sample data shared by both layers.
└── README.md

unit/ covers self-contained logic — Formulae_ math, the deck.gl and Leaflet adapters, the map engine registry, COG/STAC URL helpers, the component controller. e2e/ is currently small (smoke.spec.js, eventbus-integration.spec.js) and exercises the loaded SPA against a live backend.

How runs are wired

The root playwright.config.js is the single source of truth. Two things in it are worth knowing:

1. webServer.command is npm run start:test, which boots the Express server in NODE_ENV=test on port 8888 and waits for /api/utils/healthcheck to return 200. That is the same entry point covered in server bootstrap, just with the test environment flag set.
2. The webServer block is skipped entirely when PLAYWRIGHT_TEST_UNIT_ONLY=true is set, so unit runs do not pay the server-startup cost.

The npm scripts wrap that config:

"test":        "playwright test",                                  // everything
"test:unit":   "cross-env PLAYWRIGHT_TEST_UNIT_ONLY=true playwright test tests/unit",
"test:e2e":    "playwright test tests/e2e",
"test:headed": "playwright test --headed",                         // visible browser
"test:ui":     "playwright test --ui",                             // interactive runner
"test:debug":  "playwright test --debug",
"test:report": "playwright show-report"

Use npm run test:unit while iterating on pure logic — it is the fast loop. npm run test:e2e (or plain npm test) reuses an already-running dev server when one is up locally, otherwise spawns its own via start:test. test:headed and test:ui are the same runs with the browser visible or inside Playwright's inspector.

CI is wired through .github/workflows/playwright-tests.yml: a Postgres service container, Chromium only, two retries, with HTML reports and failure videos/screenshots uploaded as artifacts.

Fixtures

tests/fixtures/ holds plain-JS modules with frozen sample data so specs do not hand-roll geometry. coordinate-samples.js exposes coordinatePairs (short, medium, cross-continental, same-point, and Mars/Gale Crater pairs with expected distances) and bearingTestCases for compass math. geojson-samples.js exposes validGeoJSON (point, line, polygon, multi-geometries) and colorTestCases for style helpers. Both unit and e2e tests import them directly — there is no fixture-loading framework, just ES modules.